setFarValue limitations?

Abraxas) · Post by **Abraxas)** » Thu Nov 12, 2009 1:50 am

I posted previously some problems with using getRayFromScreenCoordinates, and I've found the cause of it.

I would use

camera->setFarValue(50000000.0f);

Apparently when the far value is too high, the GRFSC function produces many errors.

I'm assuming this is because of 32bit precision, am I right?

If so, what should my limits on farvalue be so then GRFSC works properly?

Thanks!

Serg88 · Post by **Serg88** » Thu Nov 12, 2009 2:53 am

float use 4 bytes

this value included in range for "long int"(signed or not) is exactly
for float must also if believe to written:
http://cplus.about.com/od/introductiont ... atdefn.htm

if you use getRayFromScreenCoordinates you also must use sentiment triangle selector.
I had same problem when i projected mouse coordinates to 3D position/

shorter: check selector

PS
also i advise you make you world smaller
use node->setscale and all fit.
50 million units is very for world
with estimated that 1 units == 1 meter in real life

Abraxas) · Post by **Abraxas)** » Thu Nov 12, 2009 2:58 am

My selector is fine.

What I did was setfarvalue to something short like 1E3, do the raycast, then set it back to 1E7 (which is what I need.. :/)

I guess another solution would be to scale everything down a hundred times.

BTW the raycast is more for plane3d::getintersectionwithline then it is for nodecollision actually, since I'm making a strategy game here.

Serg88 · Post by **Serg88** » Thu Nov 12, 2009 3:16 am

im talking about getRayFromScreenCoordinates and getCollisionPoint

especially if you make strategy don`t set so big FarValue, it`s clip far from camera polygons

hybrid · Post by **hybrid** » Thu Nov 12, 2009 9:05 am

Actually, most depth buffers are 16bit or 24bit. Using such a huge value will make the projection matrix rather wrong, and all results that involve object-to-world transformation pretty unreliable. Usual values are between 1000 and 50000.